Musical ability tests have historically had two main applications: diagnosing musical deficits and aptitudes in schoolchildren, and helping psychologists uncover the latent structure of musicality. The earliest standardised musical ability tests used pen and paper, but the development of affordable computers and new psychometric techniques has enabled the development of considerably more sophisticated ability tests.
My collaborators and I have taken advantage of these new technical opportunities and constructed a series of advanced musical ability tests. These tests are constructed with a common procedure, developed in my Master’s project under the supervision of Daniel Müllensiefen:
- Identify and formalise a relevant perceptual task, such as hearing three versions of the same melody and identifying which is the odd one out.
- Identify structural features of the task that may be manipulated to probe different ability levels (e.g. the length of the melody, the nature of the difference between melodies).
- Empirically derive a statistical model that predicts item difficulty from these different structural features.
- Use this model to generate a large and varied bank of items, each with a difficulty estimate.
- Construct a computerised adaptive test from this item bank, where item selection is automaticaly tailored to the participant’s ability level.
We have developed several tests using this approach, each of which have publicly available implementations using the psychTestR framework: